Systematic and stochastic placement error divergence between low and high-NA EUV lithography

Pattern placement error has historically been related to overlay in DUV lithography, where continuum approximations, transmissive masks, and telecentric stepper/scanners are used. EUV lithography does not enjoy such properties: stochastic phenomena - both optical and chemical - strongly affect the printed pattern even at relatively relaxed k(1) factors, while non-telecentric optical paths complicate across-focus as well as across-slit pattern behavior [1]. Both randomness and optical non-idealities contribute to unpredictable and uncorrectable feature edge misplacement as stochastic and systematic pattern placement errors, which must be taken into consideration in the overlay budget [2]. For low-NA (NA = 0.33) EUV lithography, stochastic and systematic pattern misplacement are controlled by printing very relaxed pitches, often greater than 34 nm, which correspond to a k(1) > 0.4 (in ArFi lithography production lines run at pitches approximate to 80 nm, where k(1) < 0.3). This allows for the optimization of the mask and source to control the edge placement across focus and slit. Moreover, sacrificing some throughput keeps the printed features large enough to receive enough photons to preserve the stochastic phenomena under a controllable level. In high-NA (NA = 0.55) EUV lithography arrival angles are larger - up to 33 degrees - and simulations show that this has dire consequences for pattern placement control [1]. Moreover, high-NA EUV lithography is expected to enter production for the 14A node [3], with a line/space minimum pitch of 17-19 nm (k(1) approximate to 0.37) where the edge placement error budget will be further reduced. We compare the overall pattern placement control of a well-established low-NA exposure configuration for lines/spaces at a pitch of 36 nm (k(1) = 0.44) with different high-NA solutions to print 22 nm (k(1) = 0.45) lines/spaces.